A clinical lab may check its testing process several times a day, but a problem can begin long before a result breaks a quality-control rule. Researchers at Weill Cornell Medicine in New York City have proposed an artificial intelligence system called LabARIA that could help labs spot those early warning signs.
Clinical laboratories test blood, urine, tissue and other samples to help doctors diagnose and treat illness. Their work is under growing pressure from rising test volumes, staff shortages and more complex diagnostics. When a sample is damaged or equipment drifts out of range, testing may be disrupted and results delayed. The researchers say bringing information from across lab operations together could help staff investigate sooner.
The proposed system’s full name is Laboratory Agentic Reasoning and Integrative Analytics. In a perspective published Oct. 28 in npj Health Systems, the team described LabARIA as a framework built from several specialized AI tools, called agents. A central “data coordinator” would route information from existing laboratory information systems and other databases to the agents. The plan is to connect with those systems, not replace them.
Different agents would watch different parts of the work: specimen quality, the time tests take, analytical quality control, instrument performance and compliance with regulations. Other agents would estimate future needs for supplies and staff. By linking these signals, LabARIA could notice a pattern that a tool focused on just one task might miss. For example, a gradual change tied to reagents or other testing materials might be flagged before it leads to a formal quality-control violation.
The system could also look for trouble before a sample reaches the testing stage. Specimens can be compromised during collection, handling or transport. Patterns in indicators of sample quality could prompt laboratory staff to check whether a wider problem is developing, potentially preventing interruptions and delays in reporting results.
“Laboratory medicine is a critical but largely neglected area,” said senior author Dr. Fei Wang, Weill Cornell’s associate dean for artificial intelligence and a professor of medical informatics. Lead author Dr. He Sarina Yang, an associate professor of clinical pathology and laboratory medicine and medical director of the clinical chemistry and toxicology lab at NewYork-Presbyterian/Weill Cornell Medical Center, said the system could help staff pull information together and address problems earlier.
Human oversight is central to the proposal. LabARIA is designed to keep records that show which data it used and how it reached its recommendations. Laboratory professionals would be able to review and override them. That matters in a setting where decisions affect patient care and where staff need to understand why a system raised an alert.
LabARIA is still a concept, not a tool in clinical use. Wang and colleagues plan to test it with past laboratory data and refine the framework. If those experiments provide convincing evidence, they could discuss using it in their clinical lab. They also say any system deployed would need continued monitoring for safety and effectiveness. The next step is therefore not replacing lab workers, but finding out whether AI can give them a clearer, earlier view of problems that might otherwise surface only after testing is affected.
